Agents in ECommerce

Document Sample
Agents in ECommerce Powered By Docstoc
					Agents in E-Commerce
          State of the Art
                         Authors:
          Minghua He and Ho-fung Leung
                      2002




                  Ali Fatolahi
            Presenter:

                 CSI 5389
                Winter 2006
CSI5389         www.site.uottawa.ca/~afato092   1
          Presentation Outline
•   Introduction
•   Roles of Agents in Ecommerce
•   Special Agents for Ecommerce
•   Implementation Tools/Techniques
•   Security Issues
•   Conclusion

CSI5389         www.site.uottawa.ca/~afato092   2
                       Agent
• Ongoing Execution
• Environmental
Awareness
• Agent Awareness
• Autonomy
• Adaptiveness
• Intelligence
• Mobility
• Anthropomorphism
• Reproduction
   CSI5389        www.site.uottawa.ca/~afato092   3
              Multi Agent System (MAS)
• A network of interacting agents in which
• Each agent has
   – partial information
   – limited capabilities
       • Knowledge                     Who’s in charge?!
       • Information
       • Resources
   – a limited viewpoint
• There is no global system control
• Data are decentralized


    CSI5389                 www.site.uottawa.ca/~afato092   4
Multi Agent System (MAS) Contd…
• Computation is asynchronous
• Different agents could be heterogeneous
     –    knowledge representation
     –    data format
     –    reasoning model
     –    solution evaluation criteria
     –    goal
     –    architecture
     –    algorithm
     –    language
     –    hardware platform
CSI5389                  www.site.uottawa.ca/~afato092   5
    Where to Use?
• Telecommunication Systems
• Autonomous Vehicles and Spacecraft
• Industrial Process Control
• Information Retrieval And Filtering
• User Interface Design
• E-Commerce
• Entertainment (e.g. Computer Game And Interactive
  Cinema)
• Medical Care

    CSI5389       www.site.uottawa.ca/~afato092   6
    CBB (Consumer Buying Behaviour) Model

•   Need Identification
•   Product Brokering
•   Merchant Brokering
•   Negotiation
•   Purchase And Delivery
•   Product Service And Evaluation

CSI5389          www.site.uottawa.ca/~afato092   7
          Agents in ECommerce
• B2C
     – Product Brokering
     – Merchant Brokering
     – Negotiation
• B2B
     – Supply Chain Management
• Special Techniques
     – Mobile Agents
     – Evolutionary Agents
     – Data Mining Agents
CSI5389            www.site.uottawa.ca/~afato092   8
Agents in Product Brokering (B2C)
• Agent determines what to buy?
• Main Techniques:
     – Feature-Based Filtering (e.g. Jango)
          • Looks for the keywords
     – Collaborative Filtering (e.g. Firefly)
          • Decides based on the similarities between shoppers’
            profiles
     – Constraint-Based Filtering (e.g. PersonaLogic)
          • Narrows down list of possible products based on
            user-given constraints in a step-by-step manner.
CSI5389                 www.site.uottawa.ca/~afato092          9
Agents in Merchant Brokering (B2C)
 • Agent finds the appropriate merchant for a
   given product.
 • Main technique is price comparison:
      – e.g. BargainFinder, Kasbah and Jango
 • Multi-attribute comparison is also a very
   promising technique:
      – Other value-added services must not be ignored
           • delivery time, warranty, model, make, …
      – Customer must have the option of choosing
        between price and other features of the product
      – Tete-a-Tete (T@T) is the best example.
 CSI5389                www.site.uottawa.ca/~afato092   10
      Agents in Negotiation (B2C)
• Determining the price or other terms of the
  transaction on the Internet.
• Self-Interested Agents try:
     – To maximize their own parties’ interest
     – Under bounded rationality and incomplete
       information
     – Preparing bids or offers without concern for the
       global good
• Negotiation Strategies and Protocols are
  required
CSI5389             www.site.uottawa.ca/~afato092     11
Negotiation Protocols and Strategies
 • Negotiation Protocols
      – Auctions (retail sails)
      – Contracts (task allocation and reallocation)
 • Negotiation Strategies
           •   ZERO Intelligence Strategy
           •   Heuristic Strategy
           •   Non-Cooperative Game Theory
           •   Recursive Modeling Method
           •   Bayesian Learning Method
           •   P-Strategy
           •   Possibility-Based Approaches
           •   Auctioneer's Strategies in Combinatorial Auction
 CSI5389                   www.site.uottawa.ca/~afato092          12
          Auctions and Negotiation
• Hundreds of different types of auctions
• Nowadays they’re not time-consuming any
  more
• Barriers
     – How to decide the cost values for new and
       unvalued goods.
     – How to empower agents to face with tricks
          • Rings, Lying Auctioneer, Shills and …
• Multi-attribute negotiation is still
  complicated
     – T@T tries to provide this capability
CSI5389                www.site.uottawa.ca/~afato092   13
          Contracts and Negotiation




CSI5389          www.site.uottawa.ca/~afato092   14
    Agents in B2B: Supply Chain
• A supply chain is
  – a network of business units or facilities to
       •   procure raw materials
       •   transform them into intermediate goods
       •   transform them into final products
       •   deliver the products to customers through a
           distribution system
• Primary Cost Sets in a Supply Chain
  – Inventory
  – Administrative and Information Flows
  – Production and Transportation
 CSI5389                www.site.uottawa.ca/~afato092    15
Agents in B2B: Supply Chain (Contd.)
• Supply Chain Management (SCM)
  – is the management of the activities on a supply
    chain so that
  – the products pass through the chain
       • in the shortest time with the
       • lowest cost
• SCM Focus
  – engineering efficient manufacturing processes in
    past
  – coordination of activities in the supply chain
    through knowledge management in present

  CSI5389               www.site.uottawa.ca/~afato092   16
Agents in B2B: Supply Chain (Contd.)
• SCM functions
   – coordinating the activities of each tier within the
     supply chain
   – facilitating the transaction between tiers
   – matching the supply with the market demand
• Main Roles of Agents in SCM
   – executing the routine selection process
   – negotiation about product prices
   – data transportation

  CSI5389           www.site.uottawa.ca/~afato092    17
Agents in B2B: Supply Chain (Contd.)
• Other Roles of Agents in SCM
  – monitoring the execution of the committed orders
  – informing users of some changes
  – making adjustments in a timely fashion
• Hybrid Agent Solutions are used in Practice
  – market protocol based on task dependency network
  – negotiation-based model
  – CSP Model: Constraint Satisfaction Problems
  – LCT supply chain model: Leadtime-Cost Tradeoff

  CSI5389         www.site.uottawa.ca/~afato092   18
  Special Agents for ECommerce
• Mobile Agents
   –   agents that can migrate through the netwrok
   –   real-time access to remote resources
   –   agent-to-agent negotiations
   –   Example: MAgNET
           • networked electronic trading
           • the Java mobile agent technology (aglets)
   – MAgNET
           • A buyer agent
           • A mobile agent for the buyer
           • Multiple supplier agents
 CSI5389                  www.site.uottawa.ca/~afato092   19
   Special Agents for ECommerce
• Evolutionary Agents
  – Genetic Algorithms
  – ever-improving series
  – best negotiation
    strategy
  – matching a population
  – then improve it!
  – competitive
    negotiation

 CSI5389         www.site.uottawa.ca/~afato092   20
  Special Agents for ECommerce
• Data Mining is
   – discovering new pieces of knowledge behind plenty of
     data
   – through classification, regression, summarization,
     clustering, link analysis, and sequence analysis
   – to solve business problems such as prediction and
     forecasting
• Data-Mining Agents can
   –   monitor the change of commercial data autonomously
   –   discover new findings
   –   predict some possible consequences
   –   pass the result to relevant customers or managers.
  CSI5389              www.site.uottawa.ca/~afato092        21
Implementation Tools




  CSI5389    www.site.uottawa.ca/~afato092   22
               Security Issues
• Mobile Agent Security
   – malicious hosts
   – uncooperative or untrustworthy mobile agents
   – unauthorized extraction and modification
• Host Security
   – malicious mobile agents
• Regular Protection Techniques
• Agent Security is in its Infancy

CSI5389           www.site.uottawa.ca/~afato092     23
           Security Issues (Contd.)
• Honesty of Buyers and Sellers
    – Third-Party Enforcement
          • banks, post offices and credit-cared companies
    – Unenforced Ecommerce Transactions
          • split larger exchanges into smaller parts
• Social Laws for Agents!
    – By a committee of agent-makers
• The Problem of Trust
    – Reputation Mechanisms
CSI5389                 www.site.uottawa.ca/~afato092        24
                Open Problems
• Negotiation
     – Better Protocols
     – Sophisticated Negotiations
     – Risk Assessment
• Information Discovery
     – techniques for information retrieval and
       filtering
     – growing quantity of information
• Ontology
     – specification of conceptualization
• Security
CSI5389             www.site.uottawa.ca/~afato092   25
                     Summary
• A Comprehensive Survey
     – Agents Roles in
          • B2C Ecommerce
          • B2B Ecommerce
     – Implementation Technologies
     – Security Issues
     – Open Problems
• Some agent systems such as Kasbah are not
  introduce very well.
• Links are not valid any more!
CSI5389              www.site.uottawa.ca/~afato092   26
                My Judgment
• No impressive advancement!
• Agents technology is still academic more
  than industrial
• Implemented agents rarely use well-known
  intelligence techniques. They are based on
     – Autonomy
     – Mobility
• Ideas
     – Techniques from expert systems (Negotiation)
     – What about text mining? (Discovery)
CSI5389            www.site.uottawa.ca/~afato092      27
                My Main Idea
• Major problem is trust
     – Agent to Agent
     – Human to Agent
     – Agent to Human
• An abstract trust framework
• What about the trust architecture?
     – Trust Components
     – Relationships Amongst Components
     – Specification of a Trustable Component
CSI5389            www.site.uottawa.ca/~afato092   28
          Thank You, Any Question?




CSI5389          www.site.uottawa.ca/~afato092   29

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:13
posted:7/21/2011
language:English
pages:29